Targeting Phospholipase D in Cancer, Infection and Neurodegenerative Disorders

Home , Enzyme Commission number, Phospholipase D1, PLD3

REVIEWS

Targeting phospholipase D in cancer, infection and neurodegenerative disorders

H. Alex Brown1–4, Paul G. Thomas5 and Craig W. Lindsley1–3,6 Abstract | Lipid second messengers have essential roles in cellular function and contribute to the molecular mechanisms that underlie inflammation, malignant transformation, invasiveness, neurodegenerative disorders, and infectious and other pathophysiological processes. The phospholipase D (PLD) isoenzymes PLD1 and PLD2 are one of the major sources of signal-activated phosphatidic acid (PtdOH) generation downstream of a variety of cell-surface receptors, including G protein-coupled receptors (GPCRs), receptor tyrosine kinases (RTKs) and integrins. Recent advances in the development of isoenzyme-selective PLD inhibitors and in molecular genetics have suggested that PLD isoenzymes in mammalian cells and pathogenic organisms may be valuable targets for the treatment of several human diseases. Isoenzyme-selective inhibitors have revealed complex inter-relationships between PtdOH biosynthetic pathways and the role of PtdOH in pathophysiology. PLD enzymes were once thought to be undruggable owing to the ubiquitous nature of PtdOH in cell signalling and concerns that inhibitors would be too toxic for use in humans. However, recent promising discoveries suggest that small-molecule isoenzyme- selective inhibitors may provide novel compounds for a unique approach to the treatment of cancers, neurodegenerative disorders and other afflictions of the central nervous system, and potentially serve as broad-spectrum antiviral and antimicrobial therapeutics.

Phospholipase D Phospholipase D (PLD; KEGG enzyme commission converge. They are known to participate in cellular func- (PLD). A class of number 3.1.4.4) enzymes are phosphodiesterases that tions that require membrane remodelling or biogenesis, phosphodiesterases that serve as key components of multiple signalling and such as vesicular transport, endocytosis, degranulation hydrolyse phosphatidylcholine metabolic pathways. They are encoded by a super- and cell cycle progression. The substrate for PLD1 and and other amine-containing 1 glycerophospholipids to family of genes and can be defined by several highly PLD2 is typically phosphatidylcholine, but the enzymes generate phosphatidic acid conserved motifs. These enzymes catalyse the removal are also able to hydrolyse other amine-containing and a free head group. of head groups from glycerophospholipids to generate glycerophospholipids, including phosphatidyletha- phosphatidic acid (PtdOH), a reaction that results in the nolamine, phosphatidylserine and, to a lesser extent, stoichiometric release of the free head group1–7. One of phosphatidylglycerol. the four subgroups of PLD enzymes is characterized by Many HKD motif-containing PLD enzymes also

a conserved H-X-K-X4-D-X6-G-(G/S) catalytic motif catalyse an alternative reaction to hydrolysis (that is, that is commonly known as an HKD motif. Members of transphosphatidylation), in which short-chain primary this subgroup hydrolyse phosphodiester bonds via the alcohols compete with water as a nucleophile, gener- HKD catalytic motif using a generally similar reaction ating a phosphatidyl alcohol product, such as phos- mechanism; however, some family members also exhibit phatidylbutanol (PtdBuOH) or phosphatidylethanol 1Department of lipid hydrolase activity, whereas others do not. In addi- (PtdEtOH). This alcohol-mediated transphosphatidyl- Pharmacology, Vanderbilt (FIG. 1) University School of Medicine, tion, several PLD enzymes that lack HKD motifs have ation reaction uses physiological substrates and 5 Nashville, Tennessee been described that also generate PtdOH . has catalysis rates comparable to those of hydrolysis. In 37232–6600, USA. In mammalian cells, the HKD-containing iso some cases, the phosphatidyl alcohol products mimic Correspondence to H.A.B. enzymes PLD1 and PLD2, which share highly conserved PtdOH binding to downstream targets, thereby acti- [email protected] phox and pleckstrin homology (PX–PH) domains, are vating some signalling pathways downstream of PLD 5 doi:10.1038/nrd.2016.252 almost ubiquitous . These two isoenzymes frequently enzymes, while blocking others. Erroneously, primary Published online 17 Feb 2017 serve as nodes at points where signalling pathways alcohols have widely been referred to as PLD ‘inhibitors’

REVIEWS

Author addresses kinase (DGK) signalling. From this perspective, one or both of the isoenzymes may be involved both temporally 1 Department of Pharmacology, Vanderbilt University School of Medicine, Nashville, and spatially in a specific cellular function, but PtdOH Tennessee 37232–6600, USA. depletion can be compensated for if an isoenzyme is inac- 2The Vanderbilt Institute of Chemical Biology, 896 Preston Building, Vanderbilt University, tivated. If the causal misregulation step is due to a signal- Nashville, Tennessee 37232–6304, USA. 3The Vanderbilt Ingram Cancer Center, 691 Preston Building, Vanderbilt University, ling pathway that modulates the specific PLD isoenzyme Nashville, Tennessee 37232–6838, USA. in a physiological circuit, then one of the parallel PtdOH- 4Department of Biochemistry, Vanderbilt University School of Medicine, Nashville, generating pathways may not be similarly affected. As Tennessee 37232–6600, USA. an example, purinergic receptors can induce PtdOH 5Department of Immunology, St. Jude Children’s Research Hospital, Memphis, Tennessee production via both PLD and PLC–DGK signalling in 38105–3678, USA. astrocytomas11. These pathways are interconnected and 6Vanderbilt Center for Neuroscience Drug Discovery and Department of Chemistry, seem to constitutively cross-regulate one another in the Vanderbilt University, Nashville, Tennessee 37232, USA. production of primary PtdOH lipid products as well as that of downstream products such as diacylglycerol in publications, and it is likely that some roles previously (DAG). When one pathway is blocked the other com- ascribed to PLD enzymes in studies that used alcohols as pensates, thereby ensuring the continued production of ‘inhibitors’ are really attributable to nonspecific effects essential lipid signalling products. However, the resulting and should be re-examined2. Details of the sequence molecular species differ in acyl composition depending homology among members of the PLD superfamily, and on whether they were generated via the PLD pathway or the enzymology, signalling and functions of respective the PLC–DGK pathway. In some cases, acyl composition PLD proteins, have been reviewed previously 3–6. may be important for the activation of a specific target. Recently, theoretical work was presented that For example, a recent report showed that only unsatu- describes the possible mechanisms underlying the cat- rated fatty acids containing PtdOH dissociate the dishev- alytic activity of HKD motif-containing PLD enzymes elled, EGL‑10 and pleckstrin (DEP) domain-containing using computational methods and models that are mTOR-interacting protein (DEPTOR) from mechanistic based on reaction kinetics, thermodynamics and quan- target of rapamycin complex 1 (mTORC1)12. This find- titative insights from studies of the Streptomyces spp. ing extends previous suggestions that PLD enzymes may 7 strain PMF PLD enzyme (PLDPMF) . The mechanism have a primary role in the regulation of mTORC1 and of catalytic activity includes the following steps: first, that other enzymatic sources of PtdOH have a secondary the formation of a five-coordinate phosphohistidine role. By extension, if an individual isoform of PLD is dys- intermediate and initial phosphoryl transfer during functional and thus contributes to a disease process, then which the head group is cleaved; second, the hydrolysis it is likely that isoform-specific small-molecule inhibi- of the phosphohistidine intermediate and bond disso- tors would have less adverse consequences than would ciation of the hydrolysed substrate; and third, the for- pan-reactive inhibitors. mation of a thermodynamically stable four-coordinate Interest in targeting PLD isoenzymes with small- 7 Phosphatidic acid phosphohistidine intermediate . These individual steps molecule inhibitors has steadily grown since PLD (PtdOH). A molecule that are highly conserved among enzymes that contain the family members were implicated in a variety of human functions as both a lipid HKD motif, which supports speculation that the large diseases, ranging from cancer to neurodegeneration and second messenger and a key number of highly diverse PLD enzymes evolved as a con- viral infections. Moreover, viruses such as human cyto intermediate in glycerophos- sequence of differences in the mechanism of regulation megalovirus (HCMV)13, influenza virus14 and HIV‑1 pholipid metabolism. The (REF. 15) phospholipase D pathway is by constituents of distinct cell signalling and metabolic have been shown to use PtdOH in various one pathway that is used to pathways to fulfil a variety of cellular functions for which aspects of cellular entry, intracellular trafficking and the generate receptor-mediated PtdOH generation is vital. disruption of innate immune responses. PtdOH. A major pathway of Despite their essential roles in cellular function, Understanding the mechanism of action of PtdOH metabolism is conversion to diacylglycerol by it is curious that one or more PLD isoenzymes can be the HKD motif-containing superfamily, as well as the lipid phosphatases. genetically ablated in mice without causing lethality or roles of specific PLD isoforms in cellular signalling even overt severe morbidity phenotypes. However, some metabolism and cellular function, is paramount to Transphosphatidylation isoform-specific functions have been identified in these the development of optimal therapeutic compounds. An alternative reaction in which models8–10. It can be postulated that PtdOH production Here, we provide an overview of the development of a short-chain alcohol competes with water as a nucleophile to is so indispensable to cellular function and organism isoenzyme-preferring small-molecule inhibitors of PLD generate a phosphatidyl survival that multiple redundant pathways have evolved enzymes, discuss the role of PLD family members in alcohol product instead of to prevent the global loss of PtdOH. Both metabolic infectious diseases, cancers and central nervous system phosphatidic acid. This and signalling pathways depend on PtdOH generation, (CNS) disorders, and provide an outlook on possible naturally occurring reaction, in which alcohols function as from the initial steps of the de novo glycerophospholipid applications of isoformspecific PLD inhibitors. alternative substrates of biosynthetic pathway (that is, the Kennedy pathway) to phospholipase D (PLD) signalling pathways that are initiated by specific classes PLD isoenzyme inhibitors enzymes, has an inhibitory of GPCRs, RTKs and integrins, which lead to the rapid Until 2009, few chemical tools existed to directly study effect by diverting the primary generation of signalling pools of PtdOH5 (BOX 1; FIG. 2). and thereby elucidate the function of PLD enzymes, and catalytic reaction to form new reaction products, but alcohols In mammalian cells, GPCR signal-activated PtdOH gen- no small molecules were available to dissect the indi- themselves are not inhibitors of eration occurs predominantly via the PLD1 and PLD2 vidual roles of the two main mammalian PLD isoen- PLD enzymes. isoenzymes, and phospholipase C (PLC)–diacylglycerol zymes (that is, PLD1 and PLD2) or other therapeutically

REVIEWS

a b PLD O Infectious O R * diseases 1 O R * O 2 O X P O O Stroke O– and other cardiovascular PLD Cancers Phospholipid substrate + H 0 + BuOH diseases 2 Hydrolysis Transphosphatidylation PLD inhibitors Neurodegenerative diseases DGK

O O O

O R1* O R1* O R1* O R2* O R2* PLD inhibitors O O R2* O HO O P O Bu P O O O O O– HO O– PtdOH DAG PtdBuOH

LPP PLC Figure 1 | Phospholipase D enzymes as therapeutic targets and their mechanism of action. a | Recent findings have implicated phospholipase D (PLD) enzymes as therapeutic targets in a variety of human diseases.b | Most PLD enzymes mediate both a hydrolysis reaction that generates phosphatidic acid (PtdOH) directly andNature a transphosphatidylation Reviews | Drug Discovery reaction in which primary alcohols serve as alternative substrates for the generation of a phosphatidyl alcohol lipid product. Allosteric small-molecule inhibitors block both reactions. PtdOH is metabolized to diacylglycerol (DAG) by lipid phosphate phosphatase (LPP) enzymes. PtdOH species are also generated downstream of PLC enzymes, which directly produce DAG; subsequent phosphorylation of DAG by DAG kinases (DGKs) generates PtdOH. The mechanism of transphosphatidylation has been reviewed in detail elsewhere5. BuOH, butanol; PtdBuOH, phosphatidylbutanol. *denotes long-chain fatty acid residues.

relevant PLD enzymes, such as the endocannabinoid n‑butanol block PLD-catalysed PtdOH production by regulator N‑acyl phosphatidylethanolamine-specific competing with water as a nucleophile. However, the phospholipase D (NAPE-PLD) or bacterial PldA3,5,16–20. blockade of PtdOH production is frequently incomplete Historically, the field relied on the overexpression of under experimental conditions, and thus caution must catalytically active or inactive forms of the enzymes in be exercised in interpreting data from studies that have question, or on blocking their expression using RNA- used primary alcohols such as n‑butanol 1,3,5,16. mediated interference (RNAi); however, these biochem- After decades of ‘false starts’, a brief report was pub- ical and genetic approaches did little to garner attention lished in 2007 in which a group at Novartis described from the pharmaceutical industry3,5,16–18. halopemide (FIG. 3b, compound 8; concentration that

For PLD inhibition to emerge as a viable therapeutic inhibits PLD2 activity by 50% (IC50) = 1.5 μM) and a set approach, target validation with selective small mole- of 14 analogues as PLD2 inhibitors. These analogues cules was required. Although early efforts throughout included an indole amide congener that is now referred the 2000s did identify small molecules that were capa- to as 5‑fluoro‑2‑indolyl des-chlorohalopemide (FIPI; FIG. 3b 21 ble of modulating the function of some PLD enzymes— , compound 9; PLD2 IC50 = 200 nM) . This man- both indirectly (for example, polyphenolic natural uscript only reported PLD2 activity, with no mention products such as resveratrol (FIG. 3a, compound 1) or of PLD1 inhibition, which prompted further pharmaco steroids such as triptolide (FIG. 3a, compound 2)) and logical evaluation of these compounds in both cell-based directly (phosphate mimetics such as tungstate (FIG. 3a, and biochemical assays with the purified PLD1 and compound 3) and vanadate (FIG. 3a, compound 4), as PLD2 enzymes16. Interestingly, halopemide potently FIG. 3a well as natural products such as SCH420789 ( , inhibited both PLD1 (cellular IC50 = 21 nM; biochem- FIG. 3a compound 5) and calphostin C ( , compound 6)) ical IC50 = 220 nM) and PLD2 (cellular IC50 = 300 nM;

— these ligands failed to accelerate development in the biochemical IC50 = 310 nM), as does FIPI (PLD1 cellu-

field in part owing to their obvious lack of specificity. lar IC50 = 1 nM; PLD1 biochemical IC50 = 9.5 nM; PLD2

Thus, over the past 30 years, the most commonly used cellular IC50 = 44 nM; PLD2 biochemical IC50 = 17 nM); class of molecules to study the function of PLD enzymes therefore, halopemide and the reported analogues21 are was primary alcohols (for example, n‑butanol (FIG. 3a, more accurately described as dual PLD1–PLD2 inhibi- compound 7)). As emphasized above, alcohols such as tors. Although isoenzyme selectivity remained elusive,

REVIEWS

16 Box 1 | Upstream activators of phospholipase D enzymes PLD enzymes . Through a combination of diversity-oriented synthesis, matrix libraries and subsequent Phospholipase D (PLD) activity is activated by G protein-coupled receptors (GPCRs), iterative parallel synthesis, more than 1,000 analogues 5 receptor tyrosine kinases (RTKs) and integrin receptors, among other receptors . were prepared and assayed, leading to a series of the The precise sequence of events that occur following ligand binding to cell-surface first highly PLD1‑selective inhibitors, which are exem- receptors seem to differ depending on the specific receptor and cell type3,5,102, and FIG. 3b 10 many of the pathway components in PLD activation are still undefined. plified by VU0359595 ( , compound ; PLD1 Two of the major regulators of PLD activity were initially discovered when membrane IC50 = 3.7 nM); VU0359595 has more than 1,700‑fold versus cytosolic elements were reconstituted using a novel exogenous substrate assay. selectivity for PLD1 versus PLD2 (REFS 16,25) (FIG. 3). In The monomeric GTP-binding protein ARF and phosphatidylinositol 4,5‑bisphosphate this instance, the addition of a (S)-methyl group to the 103 (PI(4,5)P2) were found to be stimulators of PLD activity in extracts from HL60 cells and ethyl linker increased selectivity for PLD1 versus PLD2 104 porcine brain . Based on this observation, it was postulated that PI(4,5)P2might even by more than 150‑fold across multiple examples in this be an essential cofactor for the PLD enzymes. However, the ability to partially activate sub-series by both diminishing activity at PLD2 and PLD activity using other anionic lipids, such as phosphatidylserine, in place of PI(4,5)P 2 enhancing inhibitory activity at PLD1 (FIG. 4). By virtue 105 was subsequently demonstrated . of the (S)-methyl group, a multitude of the resulting PLD was also shown to have multiple sites of interaction for polyphosphatidylinositol analogues displayed more than 200‑fold selectivity for lipid species106. Another report independently showed that ARF stimulated PLD activity in experiments that involved purifying cytosolic fractions and adding them to PLD1 over PLD2 within a traditionally PLD2‑preferring 16,25 permeabilized HL60 cells107. Soon after, members of the RHO GTPase family108–110 and scaffold . This unanticipated finding is a striking classical isoforms of protein kinase C (PKC)111–113 were shown to directly stimulate PLD1 example of the ‘magic methyl’ effect, and this modifica- activity. An intriguing regulatory loop was discovered when ARF6 was shown to directly tion to the linker promotes a U-shaped ligand topology. modulate phosphatidylinositol 4‑phosphate 5‑kinase (PI5K), which catalyses the Importantly, VU0359595 had markedly improved ancil- 114 rate-limiting step in the generation of PI(4,5)P2 in mammalian cells . ARF seems to lary pharmacology relative to halopemide and FIPI, and function as a feedforward stimulator of PI(4,5)P2 production, triggering the production displayed an attractive DMPK profile, which enabled of PtdOH by PLD, which requires PI(4,5)P . PtdOH then stimulates PI5K activity, which 16,25 2 in vivo proof-of-concept studies . generates intracellular PI(4,5)P . 2 Within the piperidine benzimidazolone core analogues, preference for PLD2 was elusive; thus, a number of bioisosteres of the piperidine benzimidazolone, as this discovery was an informative lead. Halopemide well as alternative GPCR-privileged structures, were sur- was originally developed by Janssen in the 1980s as an veyed26–30. The structure–activity relationship (SARs) of atypical antipsychotic that has potent antagonistic activ- these compounds were shallow, and more than 99% of the

ity at D2 dopamine receptors (as well as at more than variants studied were devoid of PLD-inhibitory activity; 30 additional biogenic amine receptors) and was called however, an N‑phenyl triazaspirone congener (FIG. 5, com- a “psychic energizer” (REFS 22,23). In clinical trials, halo- pound 11) displayed a modest preference for PLD2 and pemide was shown to have effects on the negative as well represented a considerable departure from the halopem- as the positive symptoms of schizophrenia, without the ide-based PLD inhibitors26. A subsequent lead optimiza- extrapyramidal side effects that are common to stand- tion campaign focusing on the N‑phenyl triazaspirone ard atypical antipsychotic agents22,23. Importantly, halo- congener encountered a shallow SAR, yet the addition pemide achieved high plasma exposure and displayed of a fluorine atom in the three‑position of the N‑phenyl no adverse events or abnormal biochemistry. At these moiety provided VU0364739 (FIG. 5, compound 12), the

exposures, halopemide fully inhibited both PLD1 and first highly selective PLD2 inhibitor (PLD2 IC50 = 20 nM; PLD2, which suggests that the mechanism of inhibi- 75‑fold selective for PLD2 versus PLD1)27. Moreover, tion of PLD1 and PLD2 by this chemotype would be similarly to VU0359595, VU0364739 was a direct PLD safe in humans; this represented a major step towards inhibitor, and it had potent activity in biochemical assays de‑risking this therapeutic approach. However, halo- with purified PLD proteins. In addition, VU0364739 pemide was not a panacea. Many laboratories began to had improved ancillary pharmacology relative to halo- use both halopemide and FIPI as key proof-of-concept pemide and FIPI, and a DMPK profile that was suita- tools for PLD1 and PLD2 target validation, which led to ble for in vivo proof-of-concept studies27. Interestingly, potentially flawed conclusions, as both compounds dis- PLD1‑preferring VU0359595 was peripherally restricted, play a considerable amount of ancillary pharmacology whereas PLD2‑preferring VU0364739 was CNS pene- 25–27 at a multitude of biogenic amine receptors, as expected trant (brain-to-plasma ratio (Kp) = 0.7) . An intriguing for atypical antipsychotics that have GPCR-privileged observation within this novel PLD2‑preferring series structures as a core motif 22–30. relates to the effect of the (S)-methyl group, which had Thus, we initiated an optimization programme a marked impact on PLD1 selectivity in the benzimida- that used halopemide as one starting point to develop zolone series. In the benzimidazolone series, the addition isoenzyme-selective PLD inhibitors that had improved of a (S)-methyl group into N‑phenyl triazaspirone conge- ancillary pharmacology and drug metabolism and phar- ner converted an approximately tenfold PLD2‑selective macokinetics (DMPK) profiles, and were suitable for inhibitor into the dual PLD1–PLD2 inhibitor compound Drug metabolism and in vivo proof-of-concept studies16. The design and opti- 13 (FIG. 5), which was approximately sixfold selective for pharmacokinetics mization strategy was based on results from two types PLD1 versus PLD2. These subtle structural features that (DMPK). Refers to all aspects of the absorption, of in vitro catalytic assays that were conducted in paral- promote substantial changes in PLD isoform selectiv- distribution, metabolism and lel: a cell-based assay and a biochemical assay that pre- ity are reminiscent of the ‘molecular switches’ that are excretion of a drug. sented defined substrates to human or rat recombinant prevalent in allosteric GPCR ligands25–28.

REVIEWS

Glucose Glycolysis Glycerol FA PI

O G3P dehydrogenase O GK O OH P P HO O OH HO O OH OH OH MGL Dihydroxyacetone phosphate Glycerol-3-phosphate PI kinase

FA Acyl-CoA

PLA GPAT 1–4 PIP

R1 O R1 O R1 O MGK ATX O O O O O PIP kinase P N O O + P HO – O OH LPP OH O HO OH HO Lysophosphatidylcholine Lysophosphatidic acid Monoacylglycerol

FA Acyl-CoA FA Acyl-CoA FA Acyl-CoA R2 R1 O O PLA LPCAT PLA LPAAT PLA MGAT O O

R1 O R1 O HO P O R1 O O OH O O DGK O O O PLD PLC P OH P N O OH OH OH O O + O OH O O O– LPP O O PO3H2 R O R2 O R2 O 2 H2O3P Phosphatidylcholine Phosphatidic acid Diacylglycerol PI(4,5)P2 Figure 2 | Metabolic pathways that lead to the generation of phosphatidic acid. Metabolic pathways that lead to the generation of phosphatidic acid (PtdOH) include de novo biosynthesis as well as the phospholipase D (PLD) and Nature Reviews | Drug Discovery PLC–diacylglycerol kinase (DGK)-mediated signalling pathways. The generation of PtdOH seems to be broadly exploited by pathogens as part of the infection process and thus represents a novel therapeutic opportunity; the design of novel small-molecule inhibitors may lead to new treatments for infection. The figure illustrates three routes of cellular PtdOH generation via the indicated enzyme-mediated pathways. ATX, autotaxin; FA, fatty acid; G3P, glycerol‑3‑phosphate; GK, glycerol kinase; GPAT, glycerol‑3‑phosphate acyltransferase; LPP, lipid phosphate phosphatase; LPAAT, lyso-PtdOH acyltransferase; LPCAT, lysophosphatidylcholine acyltransferase; MGAT, monoacylglycerol acyltransferase; MGK, monoacylglycerol kinase; MGL, monoacylglycerol lipase; PI, phosphatidylin-

ositol; PIP, PI phosphate; PI(4,5)P2, PI 4,5‑bisphosphate; R1, R2, fatty acyl moieties.

The unprecedented level of PLD isoenzyme selec- compounds were allosteric16,25–30. Using a backscattering tivity and the presence of apparent ‘molecular switches’ interferometry technique33, it was found that VU0359595 led us to question the mechanism by which these com- binds to two sites on the PLD1 protein (one with high pounds inhibit PLD enzymes. These features, along affinity and the other with low affinity)34, which further with the structural similarity of these compounds to supports an allosteric mode of inhibition akin to that of previously reported piperidine benzimidazolone-based allosteric AKT inhibitors31,32. Of course, in the absence allosteric inhibitors of AKT31 — that is, they have a two- of high-resolution structural determinations, the sites of site binding mode within both the catalytic domain and interaction are somewhat speculative, but some sugges- the pleckstrin homology domain — led us to postulate tions have been reported. In 2015, evidence was pub- that VU0359595 and VU0364739 were allosteric inhib- lished for two sites of inhibitor action that involved both itors31,32. Initial studies with halopemide and FIPI in the enzyme catalytic centre and an allosteric phospho assays using a PLD1.d311 truncation construct (which inositide-binding pocket35. In this study, FIPI was

Phosphatidylinositol lacked a portion of the pleckstrin homology domain) shown to interact with S757 of PLD2 (and also other 16 4,5‑bisphosphate showed a slight reduction of PLD activity ; however, PLD isoenzymes); by contrast, VU0364739 interacted

(PI(4,5)P2). A glycerophospho- both VU0359595 and VU0364739 significantly reduced with two different PLD2 sites, with one in the HKD (at lipid that is generally found in PLD activity (by more than tenfold), which suggests that S757 or S648) and the other an allosteric site (R210 or low levels within cells and the pleckstrin homology domain (a domain that is far R212) that is natively occupied by phosphatidylinositol functions as a lipid messenger 4,5‑bisphosphate as well as a substrate for other away from the HKD catalytic active site) is key for the (PI(4,5)P2). Binding of VU0364739 to second messengers in many binding of these molecules. Additional enzyme kinetic the allosteric site blocked PI(4,5)P2 binding. In combi- cell-signalling pathways. studies further increased our confidence that these nation with F244‑L245‑L246, the PI(4,5)P2-binding site

REVIEWS

a Indirect O OH O HO OH O O H OH O 1 Resveratrol 2 Triptolide

Direct OH O O O

O O O O O O O O O O O -OWO- O OH V – HOOC O O O CHO OH O OH 3 Tungstate 4 Vanadate 5 SCH420789 6 Calphostin C 7 n-butanol

b O HN N O N Cl N H F 8 Halopemide H >30 off-target N PLD1 IC50 =21 nM O N Ph PLD2 IC = 300 nM activities 50 O HN N O HN Diversity-oriented synthesis 7 off-target F N • Iterative parallel synthesis 10 VU0359595 activities O • ‘Magic methyl’ effect Br N PLD1 IC50 =3.7 nM • Triazaspirone for PLD2 N PLD2 IC50 =6,400 nM H N 9 FIPI H >30 off-target PLD1 IC50 =1 nM activities PLD2 IC50 =44 nM Figure 3 | Inhibitors of phospholipase D enzymes. a | Representative early indirect and direct inhibitors of phospholipase D (PLD). | Second-generation PLD inhibitors that are based on halopemide (providing the first b Nature Reviews | Drug Discovery isoenzyme-selective PLD1 inhibitors). IC50, concentration that inhibits PLD activity by 50%; FIPI, 5‑fluoro‑2 indolyl des-chlorohalopemide.

forms a hydrophobic pocket in the pleckstrin homol- selective for PLD2 (IC50 = 355 nM) and has negligible

ogy domain that is identical to that of allosteric AKT activity against PLD1 (IC50 ≈20 μM). Incorporation of the inhibitors31, and thus promotes a high degree of PLD2 (S)-methyl group used in the piperidine benzimidazolone selectivity35. Not only have these second-generation, series had a substantial impact on enhancing PLD1 inhib- halopemide-based and triazaspirone-based PLD inhibitory activity (increasing it by more than 250‑fold) and itors achieved high selectivity for the individual PLD1 provided ML299 (FIG. 6, compound 15), which is a potent,

and PLD2 isoenzymes, they have also provided a novel dual PLD1–PLD2 inhibitor (PLD1 IC50 = 6 nM; PLD2 16,25–30 28 allosteric mechanism for PLD inhibition . IC50 = 20 nM) . Once again, ancillary pharmacology and PLD2 has recently emerged as a target in oncology and physiochemical properties were improved, and interest- infectious diseases14,16,28,29,36, and VU0364739 has served ingly, ML298 was found to be peripherally restricted, 28 28 as a proof-of-concept compound ; however, a better tool whereas ML299 was CNS penetrant (Kp = 0.44) . was required. Although VU0364739 is 75‑fold selective Based on the ability of the triazaspirone core to pro- for PLD2 over PLD1, the selectivity is driven by potency vide a range of PLD pharmacological profiles, a variety

at PLD2 (IC50 = 20 nM), but the compound still inhibits of non‑N‑Ar (aromatic) moieties that are not bound to

PLD1 at modest concentrations (IC50 = 1500 nM); there- nitrogen were evaluated in an attempt to develop a PLD2 fore, a third-generation optimization campaign was inhibitor that was comparable to ML298 but was CNS launched27. This effort led to the synthesis of ML298 penetrant28. This effort identified ML395 (FIG. 6, com- (FIG. 6, compound 14), which is more than 53‑fold pound 16), which is a pyridylmethyl congener with good

REVIEWS

HN N N NH X R1 O

1 1 X R PLD1 IC50 PLD2 IC50 Fold PLD1- X R PLD1 IC50 PLD2 IC50 Fold PLD1- (nM) (nM) selective (nM) (nM) selective

H 260 9,000 35 5-Br 3.7 6,400 1,700 F F H 20 1,600 80 4-F 66 13,000 200 F

H 90 14,000 155 4-F 43 12,000 280 Cl

5-Cl 11 1,800 163 4-F 130 10,000 77 F F 5-Cl 6.4 1,200 190 4-F 38 14,000 370

Cl F

5-Cl 10 1,400 140 4-F 100 >20,000 >200 F

5-Cl 8 1,150 140 5-F 11 3,100 180 F F 5-Cl 3 730 240 5-F 7.4 1,040 140

F Cl

5-Br 5.5 3,900 700 6-F 2 360 180

5-Br 4 890 222 6-F 12 3,800 320

Cl F 5-Br 3.5 187 53

F Figure 4 | Representative examples of phospholipase D1‑preferring inhibitors with a benzimidazolone scaffold. Nature Reviews | Drug Discovery The phospholipase D 1 (PLD1) IC50 values (concentration that inhibits PLD activity by 50%) were determined in cellular

PLD1 assay with Calu-1 cells. The PLD2 IC50 values were determined in a cellular assay with HEK293-GFP PLD2 cells. Each 1 IC50 was determined in triplicate measurements. X denotes H or halogen, and R denotes the aryl- or cyclopropylaryl moieties (as listed in the table).

18 PLD2 potency (IC50 = 380 nM), no measurable activity PldA and the endocannabinoid regulator NAPE-PLD .

at PLD1 (IC50 > 30 μM), exceptional physiochemical Importantly, no ligands existed that could modulate properties and an excellent DMPK profile, including these structurally divergent PLD enzymes, thus hinder- 29 high CNS penetration (Kp = 1.48) . As a result, a series ing our understanding of their potential as therapeutic of potent and isoenzyme-selective, as well as dual, PLD agents. The search for inhibitors of these PLD enzymes inhibitors were developed; these compounds had DMPK led to the re-examination of certain selective oestrogen and ancillary pharmacology profiles that were suitable receptor modulators (SERMs), which have been used in for in vivo proof-of-concept experiments aimed at dis- the successful treatment of oestrogen-receptor-positive secting the roles of the individual mammalian PLD breast cancer37,38. A previous report suggested a potential enzymes (described below)16,25–30. ER-independent mechanism for the efficacy of SERMs Although a major advance in the PLD field, these such as raloxifene (FIG. 7, compound 17) and 4‑hydroxy- halopemide-derived and novel triazaspirone-based series tamoxifen (FIG. 7, compound 19; the active metabolite of mammalian PLD inhibitors were devoid of activ- of tamoxifen (compound 18)) that may be due to off-target ity against both the bacterial Pseudomonas aeruginosa inhibition of mammalian PLD enzymes39.

REVIEWS

profiling of desketoraloxifene against NAPE-PLD, and H O N it was found to be slightly more potent than was com- N pound 21 (NAPE-PLD IC = 58 μM). Thus, both deske- O 50 HN toraloxifene and compound 21 represented the first N 5 oﬀ-target entry into potentially universal inhibition of structurally F activities H and phylogenetically distinct PLD enzymes18. Although O N N 12 VU0364739 (JWJ) an advance, compound 21 is a lead compound (that is,

O PLD1 IC50 =1,500 nM the ‘halopemide of the mammalian chemotypes’) from HN PLD2 IC = 20 nM N 50 which to develop a more optimized compound for target validation of PldA and NAPE-PLD. The strategic use of diverse, interdisciplinary exper- 11 H imental approaches — including lipidomics, metabolom- O N PLD1 IC50 =1,000 nM N ics, phosphoproteomics, medicinal chemistry, molecular PLD2 IC50 =110 nM O HN pharmacology, biochemistry and DMPK optimization N PLD isoform — has led to the development of selective compounds selectivity that have allowed the roles of the individual mamma- 13 reversed lian PLD1 and PLD2 isoenzymes to be dissected; have

PLD1 IC50 =25 nM uncovered an allosteric mechanism of inhibition; and PLD2 IC = 140 nM 50 have provided entry into a chemotype that provides Figure 5 | Second-generation phospholipase D2‑selective and dual phospholipase potentially universal PLD inhibition. These compounds D1–phospholipase D2 inhibitors within the triazaspironeNature series. Reviews VU0364739 | Drug Discovery (also provide initial starting points for the development of known as JWJ; compound 12) emerged as an important phospholipase D2 inhibitors of PldA and NAPE-PLD. (PLD2)‑selective compound. IC50, concentration that inhibits PLD activity by 50%. The development of these compounds offers a new set of tools for analysing the role of PLD enzymes generally and that of specific PLD isoenzymes in diverse The potential of SERM chemotypes as PLD inhibitors biological conditions. Although early work (described was not fully explored until recently, in part because the above) implicated PLD enzymes in a number of nor- PLD-inhibitory activity required mid-micromolar con- mal and pathological processes, the selectivity provided centrations, and SERM activity might complicate defin- by these compounds allowed the direct interrogation itive target validation. However, these novel chemotypes of PLD1 and PLD2 in the host response to infectious might provide a novel scaffold that could achieve uni- diseases, including influenza virus and HIV infections, versal PLD inhibition across diverse PLD enzymes, and and in cancer. Although genetic approaches, including they have a binding mode that is distinct from the halo- inducible knockout systems, have the advantages of pemide-derived and triazaspirone-based series of mam- achieving complete deletion of an enzyme and its activ- malian PLD inhibitors16,25–30. Intriguingly, raloxifene, ity, highly selective compounds can be used in diverse 4‑hydroxytamoxifen and tamoxifen all showed modest cell types and animal models, including those in which inhibitory activity against the bacterial PldA enzyme, genetic ablation can be difficult. In addition, the dele- which prompted the evaluation of desketoraloxifene tion of PLD enzymes is likely to have effects distinct (FIGS 7,8, compound 20), which is a SERM that has a from those of short-term inhibition, the latter of which unique planar topology and is a stronger activator of provides a better model of therapeutic intervention. the activator protein 1 (AP‑1)-binding site of ESR1 (which encodes oestrogen receptor-α)) compared with PLD enzymes in infectious diseases that of ESR2 (which encodes oestrogen receptor-β)18. All viruses co‑opt the resources of their target host to Desketoraloxifene was a breakthrough, and had com- complete their replication cycle. This parasitism includes parable inhibitory activity against mammalian PLD1, the use of cellular processes and metabolite resource Lipidomics, metabolomics, mammalian PLD2 and bacterial PldA (FIG. 8). This unex- pools. Every stage of the virus life cycle requires host phosphoproteomics pected finding led to a new multidimensional optimi- cell cooperation, from virus entry (which often exploits A subset of systems zation campaign that focused on desketoraloxifene and host-cell trafficking pathways and endosome formation) biology-level analyses of the produced hundreds of novel analogues. Taking into to nucleic acid transport and replication, protein pro- constituents of cells and tissues that are usually conducted as account the well-established SERM pharmacophore duction, protein trafficking and modification, and finally a quantitative analysis between model, these analogues were designed to abolish SERM culminating in viral budding. Host cell energy, nucleic two conditions, such as healthy activity (FIG. 8), and they were then measured using acids, amino acids and lipids are all ingredients for the versus diseased tissue, or in vitro enzymatic assays against PLD1, PLD2, PldA and production of new virus particles. treatment with or without a 18 21 (FIG. 8) particular drug. The analysis of NAPE-PLD . This effort provided compound , PLD enzymes can influence each of these pathways. how lipids, aqueous metabolites which was devoid of the 6‑OH moiety that is crucial for Most directly, the role of PLD enzymes in generating and peptide phosphorylation oestrogen receptor binding and antiproliferative action. PtdOH makes them key components of the nascent change in response to a defined It also contains an N,N‑dimethylamino moiety, which small endosomes that are used for entry by many input provides insights into the is known to greatly reduce SERM activity in this core40, viruses, including influenza virus, respiratory syncytial mechanistic processes that 41–44 constitute a complex cellular but retained good activity against PLD1 (IC50 = 4.7 μM), virus, hepatitis C virus and coronaviruses . PtdOH function or pathophysiological PLD2 (IC50 = 7.1 μM), PldA (IC50 = 7.5 μM) and even is enriched in endosomal membranes, as it is required state. 18 5 NAPE-PLD (IC50 = 67 μM) . This finding also led to the for membrane curvature . Although not yet directly

REVIEWS

F H H O N O N N F N O O HN HN N 3 oﬀ-target N N No oﬀ-target F activities activities

14 ML298 Br 16 ML395

PLD1 IC50 =20,000 nM H PLD1 IC50 >30,000 nM O N PLD2 IC50 =355 nM N PLD2 IC50 =380 nM O HN N 3 oﬀ-target F activities

15 ML299

PLD1 IC50 =6 nM

PLD2 IC50 =320 nM Figure 6 | Third-generation phospholipase D2‑selective and dual phospholipase D1–phospholipase D2 inhibitors within the triazaspirone series. These compounds have a superior ancillary pharmacologyNature profile Reviews compared | Drug withDiscovery PLD

inhibitors 8-13 (FIGS 3,5). IC50, concentration that inhibits PLD activity by 50%; PLD, phospholipase D.

demonstrated, the inhibition of PLD enzymes could Prophylactic treatment with VU0364739 provided potentially limit the efficiency of endosome forma- significant protection in a mouse model of lethal chal- tion and provide an early block in the virus replication lenge with several influenza virus strains, including the cycle. Furthermore, the activation of PLD enzymes and pathogenic H7N9 avian influenza virus A/Anhui/1/2013 the production of PtdOH can support multiple signal (REF. 14). The low solubility of VU0364739 limited the transduction cascades, including those mediated by ability to deliver it at high doses, and the vehicle had mitogen-activated protein kinases (MAPKs; for exam- some toxicity on its own, which suggests that more robust ple, p38), which have been shown to be necessary for protection from mouse lethality might be obtained with the activation of early endosomal trafficking molecules the newer generation of inhibitors that have improved such as RAB5 and early endosome antigen 1 (EEA1)45. DMPK and higher solubility in non-toxic delivery Making use of the extensive library of available com- systems. Importantly, the compound itself showed pounds that inhibit diverse forms of PLD, studies have no added toxicity beyond the vehicle, as measured by been undertaken that dissect the roles of PLD enzymes extensive blood chemistry profiles, liver toxicity pan- in the host response to infectious diseases, including els and a modified Irwin Neurological Battery14, which influenza virus and HIV, and the role of pathogen PLD measures a wide variety of parameters in the autonomic enzymes in specific bacterial infections. nervous system and the somatomotor system. These promising results led to the investigation of the Influenza virus. The potential role of PLD enzymes mechanisms by which PLD inhibition promoted survival in crucial events in the virus life cycle led us to inves- after influenza virus infection in vivo and prevented pro- tigate the function of these enzymes during influenza ductive infection in vitro. As noted, there are multiple virus infection14,29. Indeed, following challenge of host stages of the virus life cycle that could be affected by cells with influenza virus in vitro, PLD catalytic activity PLD inhibition. Using a combination of microscopy and was markedly increased, as measured by increases in RNAi, it was demonstrated that viral entry was signif- PtdBuOH mass. Pretreatment of respiratory epithelial icantly delayed after prophylactic treatment of human cells with a primary alcohol or RNAi targeting PLD1 and A549 adenocarcinoma alveolar cells with VU0364739 PLD2 resulted in a significant decrease in the number of in vitro14. This was true for infectious viruses and for cells that were infected with influenza virus and in the non-infectious transferrin uptake, which is indicative infectious titre produced by the culture. of a general defect in endocytosis. However, the block A panel of small-molecule PLD inhibitors were tested was not complete, and some virus was eventually able in similar assays, and isoform-specific inhibitors of PLD2 to enter the cell. The maturation of the endosome — including VU0364739, ML298 and ML395 —showed was also affected, and the expression of RAB5, EEA1 potent antiviral activity. In in vitro experiments, these com- and CD63 was significantly delayed when comparing pounds inhibited cellular infection and viral replication of VU0364739‑treated cells with cells treated with the a range of clinically relevant influenza viruses, such as the vehicle control14. 2009 pandemic H1N1 influenza virus strain, the recently Importantly, this ‘mechanical block’ was not suffi- emerged avian H7N9 influenza virus strain and a repre- cient to confer protection. Rather, the host-cell innate sentative of the highly pathogenic H5N1 influenza virus immune response — which was mediated by retinoic subtype14,29. The addition of VU0364739 to the culture acid-inducible gene I (RIG‑I), interferon-regulatory medium reversed the accumulation of PtdBuOH in influ- factor 3 (IRF3) and MxA (also known as Mx1), as enza virus-infected cells, which indicates that the catalytic determined by RNAi — was required for the protective activity of PLD2 has a role in influenza virus infection14. effect of the PLD2 inhibitor14 (FIG. 9). The delay in virus

REVIEWS

15 O HIV and influenza virus by PLD inhibitors . PLD inhi- O N N bition has previously been shown to block RAS–MAPK O activation in SK‑BR3 breast cancer cells46, and in in vitro assays with CD4+ T cells, treatment with the PLD1 inhib- OH itor VU0155069 reduced extracellular signal-regulated HO S kinases (ERK) phosphorylation after T cell activation15. 17 Raloxifene 18 Tamoxifen ERK activation during T cell stimulation is required for 47 MYC induction , which, in turn, is a crucial event in PLD1 IC50 =12,000 nM PLD1: no eﬀect

PLD2 IC50 =5,000 nM PLD2: stimulator the metabolic reprogramming that occurs during T cell PldA IC = 4,000 nM PldA IC = 15,000 nM 50 50 activation. HIV targets these activated T cells and makes O use of the increased metabolic stores that are generated O 15 N N to support rapid cell division and effector activity . Consistent with this, treatment with the PLD1‑selective compound VU0359595 reduced the induction of MYC in these T cells in in vitro experiments. Events that occur

OH downstream of MYC activation, including the induction S OH HO and phosphorylation of ribosomal protein S6 kinase, 15 19 4-hydroxytamoxifen 20 Desketoraloxifene were also reduced after PLD1 inhibition . Deoxyribonucleotide triphosphate (dNTP) stores, PLD1 IC50 =6,500 nM PLD1 IC50 =6,100 nM

PLD2 IC50 =20,000 nM PLD2 IC50 =2,600 nM which are necessary for the rapid replication of T cells PldA IC = 18,000 nM PldA IC = 9,900 nM 50 50 during their expansion phase, are one of the key Figure 7 | Selective oestrogen receptor modulator chemotypes. 17–20 selective metabolic resources that accumulate after T cell activa- oestrogen receptor modulator (SERM) chemotypes haveNature inhibitory Reviews activity | Drug against Discovery tion and metabolic reprogramming, and HIV also makes 48,49 mammalian phospholipase D1 (PLD1), PLD2 and Pseudomonas aeruginosa PLD (PldA). use of these pools to complete its replication cycle . To

IC50, concentration that inhibits PLD activity by 50%. establish that the depletion of dNTP stores in activated T cells is a key mechanism downstream of PLD1 inhibition, dNTPs were added to VU0359595‑treated CD4+ T cells and shown to rescue HIV replication in cell-based endocytosis permitted a more efficient and effective assays15. Thus, PLD1 inhibition in vitro restricts HIV antiviral response, which limited subsequent productive infection by altering host-cell metabolic stores, which virus replication. This novel mechanism demonstrated disrupts the virus life cycle. the potential for broad-spectrum antiviral activity of New therapeutic compounds against HIV are increas- PLD2 inhibition and gave some indication of why the ingly essential owing to the emergence of resistant strains. treatment showed remarkably low toxicity (that is, There is also an unmet need to develop anti-HIV drugs although endocytosis was slowed, it was not blocked, that work in the CNS, where HIV has recently been shown thus permitting normal cell metabolism and cellular to localize, mutate and possibly become more resistant to processes to continue, albeit with altered kinetics). These conventional antiretrovirus cocktails50. The ability of data also suggested that virus exocytosis, as measured by some PLD inhibitors to penetrate the CNS27,28 may pro- RAB10‑mediated accumulation of surface-trafficking vide a novel approach to treating otherwise difficult- endosomes, was delayed, probably through mechanisms to‑reach reservoirs of the retrovirus. Interestingly, a path- similar to those that that limited virus entry. This effect way with a role that is strikingly similar to that of PLD1 seemed to be less striking than the entry phenotype, in HIV infection was recently described; PLD activity but continued exploration of the potential for multiple was shown to promote RNA replication in a red clover pathways of action of PLD2 inhibition on influenza virus necrotic mosaic plant virus51. This suggests that PtdOH replication is ongoing. modulation of virus replication is an evolutionarily con- In summary, PLD2 inhibition can protect against served pathway and that this signalling lipid has a broad influenza virus infection by delaying, but not blocking, role in the infectious processes of pathogens. endocytosis, which allows the host-cell innate immune FIGURE 10 depicts pathways downstream of PLD1 and system to mount an effective response against influ- PLD2 isoenzymes and illustrates their role in the infection enza virus. This ‘extended window’ of antiviral activity process with pathogenic viruses. These pathways can be can significantly reduce productive virus replication targeted for therapeutic purposes. The pathway involv- (the steps in which PtdOH contributes to infection are ing PLD2, signalling via AKT, connects to autophagic summarized in FIG. 9). flux, whereas PLD1 has recently been shown to modulate pyrimidine biosynthesis via the enzyme CAD (which con- HIV. The generalized pathway by which PLD2 inhibition tains carbamoyl aspartate synthase, aspartate transcarba- protected against influenza virus infection suggested mylase and dihydro-orotase domains), which makes CAD that similar efficacy might be seen with other enveloped a potential target for future antiretroviral therapeutics52. viruses. Supporting this hypothesis, the PLD1 inhibitor Although the pathways downstream of PLD activity VU0359595 greatly reduced the ability of HIV to repli- that can be exploited by viruses are distinct, they suggest cate in CD4+ T cells in vitro15. Surprisingly, there were a common mechanism for how inhibitors of particular distinctive elements in the mechanism of suppression of PLD isoforms might serve as broad-spectrum inhibitors

REVIEWS

Reduces O O N SERM N activity

Multidimensional 3 2ʹ 3ʹ 5 2 parallel synthesis 6-OH deletion OH 4ʹ ablates SERM S S activity HO 6 21 OMe

PLD1 IC50 =4.7 μM 20 Desketoraloxifene PLD2 IC50 =7.1 μM 7 PldA IC50 = .5 μM

NAPE-PLD IC50 =67 μM Figure 8 | Synthesis of the first universal phospholipase D inhibitor (compound 21). Chemical optimization of desketoraloxifene (compound 20) delivered compound 21, which is the first universal inhibitor of phylogenetically and Nature Reviews | Drug Discovery structurally diverse PLD enzymes that is devoid of selective oestrogen receptor modulator (SERM) activity. IC50, concentration that inhibits PLD activity by 50%; NAPE-PLD, N-acyl phosphatidylethanolamine-specific phospholipase D.

of infection. Short-term inhibition of a particular PLD NgPLD, was identified following its release into cultures enzyme can influence numerous cellular processes that of host cells that were challenged with the bacterium56. affect viral infectivity and replication efficiency, and for Although the mechanism of secretion has not been iden- many viruses, even a modest reduction in early repli- tified, NgPLD has been shown to induce the recruitment cation efficiency can completely alter the subsequent of CR3 to the cell surface and to promote cytoskeletal infectious process and allow the host to successfully rearrangement. The facilitation of infectivity by NgPLD mount an immune response14. Balancing this effect, the is known to require its interaction with a human host-cell redundancy of the lipid signalling products of the reac- factor: namely, the serine–threonine kinase AKT57. tion catalysed by PLD enzymes might explain the limited Several other bacterial pathogens have been shown in vivo toxicity of isoform-specific inhibitors. to express Pld genes that function as virulence factors, The studies described above focused on mammalian including P. aeruginosa, Legionella monocytogenes, (host) PLD enzymes as targets for disrupting pathogen Chlamydia trachomatis and Acinetobacter baumannii, replication. However, some bacterial and fungal path- among others5,58–63. P. aeruginosa is an opportunistic ogens contain their own evolutionarily distinct PLD pathogen that afflicts immunocompromised individu- enzymes, which can potentially be targeted with direct als and is a major cause of hospital-acquired infections. antimicrobials. Moreover, there is evidence that certain Patients with cystic fibrosis are particularly susceptible to parasites can exploit pathways downstream of PLDs. chronic infection with P. aeruginosa, and infection ulti- mately leads to respiratory failure. Unlike other bacterial Bacteria and parasites. The activity of PLD enzymes Pld enzymes, P. aeruginosa PLD (PldA) is comparatively in neutrophils and leukocytes is associated with anti large at 122 kDa and has a much higher homology to microbial activities such as phagocytosis, chemotaxis and eukaryotic PLD enzymes than to other bacterial iso- membrane ruffling5, but an evolutionarily older function forms, probably as a result of a horizontal gene transfer62. in bacteria may have been to promote bacterial fitness In P. aeruginosa, PldA is a secreted effector of the H2 type in the mammalian host. Several intracellular pathogens VI secretion system. Similarly to many other bacterial have genes encoding protein products that promote Pld virulence factors, PldA is internalized and induces virulence through increases in pathogen internaliza- changes in vesicle trafficking in host cells, but it has the tion or intracellular survival. These genes are gener- unusual property that its catalytic activity is stimulated ally non-essential to the survival of the bacterial strain, by polyphosphatidylinositols, which may explain the but promote entry, immune evasion or inhibition of observed localization of PldA in intracellular vesicles63. host-cell immune defences. The conventional benzimidazolone and triazaspirone PLD enzymes that function as virulence factors are series compounds have little to no effect on P. ae r ug i - most common in obligate and facultative intracellular nosa PldA and are likely to have minimal efficacy against Gram-negative bacteria, such as Yersinia pestis, which other bacterial PLD enzymes. However, a series of des is the causative agent of bubonic plague53. The Yersinia ketoraloxifene-based compounds, such as compound 21, murine toxin (Ymt) contains two HKD motifs and is that are effective inhibitors of PldA, was recently iden- highly toxic when injected into mice54. Ymt is involved in tified18 and may provide novel pharmacological inter- transmission of the pathogen to fleas and is not required ventions. The most potent compound in this study was for virulence in mice55. raloxifene, which is a SERM that had previously been The exclusively human pathogen Neisseria gonor- identified as a direct inhibitor of mammalian PLD activ- rhoeae is a Gram-negative bacterium that causes the ity39. A series of analogues in which oestrogen receptor sexually transmitted disease gonorrhoea. This bacterium binding was eliminated have been characterized (FIG. 8); invades cervical epithelial cells after initially binding to the yet analogues such as compound 21 retain the ability complement CR3 receptor, which initiates membrane ruf- to inhibit human PLD1 and human PLD2, as well as fling and endocytosis. A secreted 55 kDa PLD homologue, P. aeruginosa PldA18. This new compound series provides

REVIEWS

Budding virus (egress) Inﬂuenza virus

Cytoplasm PLD2

Autophagosome Viral assembly

PLD2 PLD2 P P P AKT Endosome P AKT

P Beclin 1 Rubicon

Virus uncoating

H+ Nucleus

H+ ions imported into the endosome acidify the interior Induction of interferons and other antiviral eﬀector elements (such as IFITM3 and MxA) Phosphatidic acid Figure 9 | Illustration of lipid involvement in host cell infection with influenza virus. The phospholipase D2 (PLD2) signalling axis is involved in both the entry and egress of influenza virus, as well as the processNature Reviewsof autophagy, | Drug which Discovery can have antiviral function. PLD activity facilitates viral entry, whereas PLD inhibition delays the kinetics of viral endocytosis14, thus widening the window of time in which the host cell can effectively prevent virus replication by producing antiviral effector molecules, such as interferon-induced transmembrane protein 3 (IFITM3) and MxA. Similarly, viral egress appears to be impaired by PLD inhibition14. The regulation of AKT activity by PLD2 has also been shown to activate the autophagy pathway, which can limit virus production. PLD-generated phosphatidic acid (PtdOH) recruits AKT to cellular membranes, and AKT subsequently phosphorylates beclin 1 at serine 295. This leads to dissociation of the beclin 1–rubicon complex, a process that promotes autophagy. PLD inhibitors reduce PtdOH concentrations, reduce AKT-mediated phosphorylation of beclin 1 and stabilize the beclin 1–rubicon complex to stall autophagy. The changes depicted in membrane lipid composition are based on previous work that investigated the contributions of PtdOH to membrane curvature115,116.

an excellent starting point for the development of more and tumorigenesis in a variety of cancer types, includ- potent and selective inhibitors of PLD enzymes. It also ing breast16,69–71, ovarian72,73, lung74, colon75,76, renal77, includes compounds that have been used in the clinic pancreatic78, prostate79 and brain cancer3,36,52, among for decades that unknowingly suppressed the activity of others. Such a widespread occurrence of elevated PLD particular PLD enzymes when used at therapeutic doses. activity and expression in human cancers suggests a Recently, it was found that certain parasites might connection between processes that are central to cellular also utilize pathways downstream of PLDs. This was transformation and PtdOH signalling. illustrated by in vitro experiments showing that PLD RTKs, which can act as oncogenic drivers in cancer, inhibitors such as VU0285655 can block the prolifera- are associated with the activation of PLD2, and there tion of Plasmodium falciparum (one of the causal agents is evidence that PLD1 is also a downstream signalling of malaria) and that of Toxoplasma gondii (which causes component of RTK-induced signalling cascades3,5. RTKs toxoplasmosis)64. such as epidermal growth factor receptor (EGFR) initi- ate signalling cascades that are known to have roles in PLD in cancer signalling pathways cellular transformation, including signalling pathways Increased expression of PLD enzymes, their subcellular that lead to the production of lipid second messengers. mislocalization and altered PLD catalytic activity have Binding of the hormone epidermal growth factor leads to been implicated as contributing factors in several types EGFR dimerization, which is followed by the autophos- of human cancer, and recent reviews have examined the phorylation of multiple tyrosine residues that form bind- role of PLD1 and PLD2 in pathways involved in cancer ing sites for Src homology 2 (SH2) domains, including progression and tumorigenesis3,65–68. Numerous reports those of growth factor receptor-bound protein 2 (GRB2), have directly linked PLD enzymes to transformation PLCγ and the p85 subunit of phosphoinositide 3‑kinase.

REVIEWS

PLD2 binds to an SH3 domain of GRB2, and this is an promotes a feedforward mechanism by which PLD2 acti- important step in the catalytic activation of PLD2 and vates RAS GTPase; this, in turn, leads to RALA GTPase in the generation of PtdOH; this step also facilitates the activation, which further modulates PLD activity3. recruitment of the guanine nucleotide-exchange protein A seminal observation was that in vivo tumour for- son‑of‑sevenless (SOS) to the membrane80. This event mation by HRAS-driven cancer cells can be blocked by

Starvation Non-starvation

PLD

Stressed conditions (PLD2) Basal conditions (PLD1)

AKT mTOR

P Beclin 1 P CAD

Autophagic ﬂux

Cellular material Pyrimidine to be degraded biosynthesis

LC3

Autophagosome

p62 p62 LC3 LC3 dNTP production

Lysosome LC3

p62 p62 LC3 LC3

LC3 LC3 LC3

Autolysosome Phosphatidic acid Phosphatidylcholine

Figure 10 | PLD isoenzyme pathways with antiviral function. Left panel: phospholipase D2 (PLD2)‑generated phosphatidic acid (PtdOH) recruits and binds to AKT and subsequently promotes autophagy via beclin 1, as detailed in FIG. 9. During the process of autophagy, autophagosomes fuse with endosomes and lysosomes, which leads to their maturation into autolysosomes. These display the autophagy markers LC3 (light chain 3, also known as MAP1LC3) and p62 (also known as sequestosome 1). This process results in acidification and allows theNature autolysosome Reviews |to Drug digest Discovery captured cellular material such as viruses. The respective details of PLD1 and PLD2 activation are discussed elsewhere3. Right panel: this shows a PLD1 pathway that affects pyrimidine biosynthesis and intracellular deoxyribonucleotide triphosphate (dNTP) levels via the enzyme CAD (which contains carbamoyl aspartate synthase, aspartate transcarbamylase and dihydro-orotase domains). PLD1 allosterically modulates the activity of mechanistic target of rapamycin (mTOR), which leads to the phosphorylation of CAD and enhances its catalytic activity. The subsequent production of pyrimidine nucleotides via the activation of CAD is independent of AKT. PLD1 inhibition, in turn, can lead to reductions in dNTP production.

REVIEWS

the co‑expression of a catalytically inactive form of PLD1 have been shown to activate PLD enzymes86–88. In the (REF. 81). This compelling report boosted the search for CNS, PLD1 and PLD2 display a distinct differential dis- isoenzyme-selective, small-molecule inhibitors of PLD tribution and function, and numerous studies in the past activity. This effect of PLD1 inhibition on HRAS-driven 10 years have begun to delineate the potential of both tumour formation probably represents just one of sev- isoenzymes to represent therapeutic targets in CNS dis- eral key roles of PLD1 in RAS-mediated oncogenesis; the orders. Previous efforts relied on the overexpression of role of PLD1 in this process was reiterated by the recent either catalytically active or inactive forms of PLD1 or observation that the novel oncogene FAM83 acts down- PLD2 in cells; on mice in which particular PLD isoen- stream of PLD catalytic activity to promote mammary zymes were genetically deleted; or on the use of RNAi epithelial cell transformation via a RAS-dependent and to knock down individual PLD isoforms in an effort to EGFR-dependent mechanism82,83. discern their discrete roles in brain disorders8,9,86–94. One of the hallmarks of cancer is the elevated use of In 2001, it was found that upon inducing experimen- glucose via glycolysis. Historically, this was thought to be tal autoimmune encephalomyelitis (EAE) in rats, PLD1 primarily in response to the higher energy needs of cancer expression increased in the spinal cord and in EAE cells relative to those of normal cells, but it is now appre- lesions95. This study suggested that PLD1 has a role in ciated that most cancer cells obtain their cellular energy autoimmune CNS inflammation, and is involved in the through oxidative phosphorylation3. Recent evidence sug- activation of macrophages and astrocytes during the for- gests that rapidly dividing cells with dramatic growth rates mation of EAE lesions, which suggests that PLD1 inhibi- use the glycolytic pathway as a means of biosynthesizing tors are possible therapeutics for the treatment of multiple lipids, amino acids and nucleotides3. These intermediate sclerosis and other neuroinflammatory diseases. PLD1 molecules may be viewed as ‘metabolic currency’, and dys- was also shown to be upregulated in the spinal cords functional regulation of signalling pathways is a frequent of rats with clip compression injuries, which suggests cause of the abnormal metabolic rates of cancer cells. This that PLD1 inhibitors might also be useful for the treat- phenomenon of higher metabolic rates of cancer cells rel- ment of spinal cord injury89. In addition, PLD1 has been ative to healthy cells and the preference of cancer cells to implicated in arterial thrombosis and ischaemic stroke8. use aerobic glycolysis is known as the Warburg effect, and In 2010, Pld1−/− mice were shown to display impaired recent insights into cancer cell metabolism have led to the α2bβ3 integrin activation and defective glycoprotein 1b‑ development of small-molecule inhibitors that disrupt dependent aggregate formation, which led to protection these cellular pathways. Both PLD1 and PLD2 have been from thrombosis and ischaemic brain injury without shown to at least partially modulate essential pathways increasing bleeding times8. This finding was recapitulated that are involved in metabolic reprogramming67. Small- in 2013 in a study that used the PLD inhibitor FIPI and molecule inhibitors of PLD isoenzymes may, therefore, demonstrated that pharmacological inhibition of PLD be better tolerated than conventional chemotherapeutic also reduced occlusive thrombosis, led to smaller infarct agents because they only partially suppress the stress- sizes, and improved motor and cognition function90. mediated responses of metabolic pathways as opposed to Among the exciting discoveries relevant to CNS dis- shutting them down completely36. orders is the identification of PLD as a potential thera- Many extracellular signals, such as hormonal stimu- peutic target in Alzheimer disease. Studies reporting that lation and acute removal of nutrients, can upregulate the PLD1 was upregulated in the mitochondrial fraction of production of certain metabolic species. Recently, it was brain tissue from patients with Alzheimer disease, and demonstrated in vitro that the PLD1‑selective inhibitor that the β‑amyloid (Aβ) region of amyloid precursor VU0359595 and the PLD2‑selective inhibitor VU0364739 protein (APP) binds to and stimulates PLD91,92, con- transiently reduce pyrimidine biosynthesis in human glio- tributed to recent interest in the development of lead blastoma-derived cells52. A detailed comparison of parallel compounds. In 2010, using Pld1−/− and Pld2−/− mice, the metabolomic and phosphoproteomic analyses suggested non-overlapping roles of PLD1 and PLD2 in the patho- that these effects were mediated through the modulation genesis of Alzheimer disease were demonstrated, along of the enzyme CAD, which acts downstream of mTOR. with the potential of therapeutically targeting these This is consistent with previous reports that demonstrated enzymes9. Overexpression and biochemical studies the modulation of mTOR pathways by PLD-generated showed that PLD1 (but not PLD2) regulates the traf- PtdOH68,84 , and may explain why some PLD inhibitors, ficking of APP and the assembly of the γ‑secretase com- via their indirect effect on CAD function, can have activity plex via a direct interaction with presenilin 1 (PS1)9,93. In against both cancer cells and viruses15,16,46 (see also FIG. 10). in vivo experiments, PLD2 was shown to be required for the synaptotoxic action of Aβ, and Pld2 ablation rescued PLD and the CNS memory deficits and promoted synaptic protection in Since the first description of mammalian PLD activity SwAPP mice, despite a high Aβ load9. Finally, in 2014, in 1973 in rat brain extracts85, the role of PLD enzymes a genome-wide association study identified a rare variant in the CNS has been of intense interest86–88. It is now in PLD3 that doubled the risk for late-onset Alzheimer appreciated that signal-dependent activation of PLD disease (LOAD); however, less than 1% of individuals enzymes is observed in a diverse array of brain-derived with LOAD carry the PLD3 mutation94. PLD3 is highly and neuron-derived cells (for example, glial cells, pri- expressed in brain regions that are prone to Alzheimer mary neurons, and glioma and neuroblastoma cells), disease, such as the cortex and hippocampus, and is and various neurotransmitters and neuromodulators expressed at significantly lower levels in neurons from

REVIEWS

the brains of patients with Alzheimer disease than in disease99 and in cancer, PLD1 inhibitors may have sim- control brains94. Unfortunately, it is unknown whether ilar mechanisms of action in these different conditions. any of the existing PLD-inhibitor chemotypes modulate Moreover, the PLD2–AKT signalling axis seems to be PLD3 catalytic activity, and the precise biological func- a common node of importance in several pathogenic pro- tion of PLD3 in the human CNS is also unknown. It is cesses. This pathway lies upstream of the beclin 1–rubicon also noteworthy that several recent reports have ques- complex36 and other key components that regulate cellular tioned some of the initial findings regarding the molec- autophagy100, illustrating a shared role of lipid metabolism ular association between rare alleles of PLD3 and the in cancer, infection and CNS conditions. occurrence of early onset Alzheimer disease96–98. Targeting distinct PLD isoenzymes may have With the next generation of potent, isoenzyme- broad therapeutic potential and it is likely that small- selective and highly brain-penetrant PLD inhibitors now molecule inhibitors of PLD isoenzymes will be tested available, the next few years should provide promising for efficacy in diseases for which there is currently advances in targeting PLD isoenzymes in multiple CNS an unsatisfactory standard of care. The absence of disorders and should identify new targets in cellular lipid toxic effects in animal models is highly encouraging. signalling cascades. Many details remain to be discovered about the roles of PLD isoenzymes, but the development of multiple Looking ahead and moving forward chemical scaffolds that function as small-molecule, Lipid signalling metabolite intermediates provide poten- isoenzyme-selective inhibitors of PLD provides an tial targets for therapeutic intervention in cancer, infec- important step in defining the therapeutic value of tar- tious diseases, and CNS disorders. Here, it is of note geting both mammalian and bacterial forms of PLD. It that the production of newly synthesized PtdOH, as will be essential to continue and extend quantitative a consequence of infection-induced activation of cellu- measurements of PtdOH and its metabolic products lar signalling pathways, may be under-reported. PtdOH in various disease states, and to identify whether the accumulation may be highly transient owing to the rapid source of disease-associated alterations in levels of flux of these lipid species, as enzymes convert PtdOH into PtdOH is the misregulation of PLD1, PLD2 or some DAG, lysophosphatidic acid (lysoPtdOH) and constitu- other enzyme that is involved in the production of ents of the Kennedy pathway (see FIG. 2). Investigations this important signalling lipid. With current advances in HCMV-infected cell lines suggested that PtdOH gen- in mass spectrometry-based metabolomic, lipidomic eration in response to infection occurred via de novo and phosphoproteomic analyses, new participants biosynthesis and not as a result of the activity of PLD in established signalling and metabolic pathways are enzymes13. However, we observed increases in PtdOH being revealed, which provide exciting opportunities mass following the infection of cell lines with influenza for therapeutic targeting. virus, an effect that was reduced after treatment with iso- A key question to be answered is whether PtdOH form-selective PLD inhibitors14. This leads us to propose predominantly serves a biophysical role (for example, the ‘PtdOH hypothesis of infection’, which postulates that in entry and egress of viruses), or whether it primarily PtdOH accumulation may be a component of infection affects nuclear transcriptional events or other innate that is broadly relevant to viral, bacterial and fungal path- immune processes that are also exploited by bacteria ogens. Moreover, although the roles of PLD enzymes in and parasites. The structural roles of lipids as com- cancer, and infectious and neurodegenerative diseases are ponents of the viral envelope also remain an area for undoubtedly complex and multifaceted, it is tempting to further scrutiny101. PtdOH generation occurring via speculate that this reflects a shared role of intermediate de novo synthesis in the Kennedy pathway, the PLD metabolism in each of these diseases. pathway and the PLC–DGK pathway needs to be fur- The recent discovery that PLD1 influences dNTP ther interrogated experimentally in other pathogenic stores through effects on mTOR and CAD15,52 provides microorganisms. This will allow to determine whether new insights into how small-molecule inhibitors of PLD1 inhibitors of these pathways could serve as broad- affect retroviral replication. Since abnormalities in mTOR spectrum anti-infectious agents, which are crucially signalling also occur in the early stages of Alzheimer needed as drug-resistant strains continue to emerge.

1. Ponting, C. P. & Kerr, I. D. A novel family of 4. Peng, X. & Frohman, M. A. Mammalian phospholipase 9. Oliveira, T. G. et al. Phospholipase D2 ablation phospholipase D homologues that includes D physiological and pathological roles. Acta. Physiol. ameliorates Alzheimer’s disease-linked synaptic phospholipid synthases and putative endonucleases: (Oxf). 204, 219–226 (2012). dysfunction and cognitive deficits. J. Neurosci. 30, identification of duplicated repeats and potential 5. Selvy, P. E., Lavieri, R. L., Lindsley, C. W. & Brown, H. A. 16419–16428 (2010). active site residues. Protein Sci. 5, 914–922 (1996). Phospholipase D: enzymology, signaling, and chemical 10. Norton, L. J. et al. PLD1, rather than PLD2, regulates This is the original report that identified the modulation. Chem. Rev. 111, 6064–6119 (2011). phorbol ester-, adhesion dependent-, and Fcγ- sequences that define the PLD superfamily. 6. Oude Weernink, P. A., Han, L., Jakobs, K. H. & receptor-stimulated reactive oxygen species 2. Sato, T., Hongu, T., Sakamoto, M., Funakoshi, Y. & Schmidt, M. Dynamic phospholipid signaling by production in neutrophils. J. Cell Sci. 124, Kanaho, Y. Molecular mechanism of N‑formyl- G protein-coupled receptors. Biochim. Biophys. Acta. 1973–1983 (2011). methionyl-leucyl-phenylalanine-induced superoxide 1768, 880–900 (2007). 11. Scott, S. A. et al. Regulation of phospholipase D generation and degranulation in mouse neutrophils: 7. DeYonker, N. J. & Webster, C. E. Phosphoryl transfers activity and phosphatidic acid production after

phospholipase D is dispensable. Mol. Cell. Biol. 33, of the phospholipase D superfamily: a quantum purinergic (P2Y6) receptor stimulation. J. Biol. Chem. 136–145 (2013). mechanical theoretical study. J. Am. Chem. Soc. 135, 288, 20477–20487 (2013). 3. Bruntz, R. C., Lindsley, C. W. & Brown, H. A. 13764–13774 (2013). 12. Yoon, M. S. et al. Rapid mitogenic regulation of the

Phospholipase D and phosphatidic acid signaling 8. Elvers, M. et al. Impaired αIIbβ3 integrin activation and mTORC1 inhibitor, DEPTOR, by phosphatidic acid. pathways as targets for therapeutics: roles in cancer. shear-dependent thrombus formation in mice lacking Mol. Cell 58, 549–556 Pharmacol. Rev. 66, 1033–1079 (2014). phospholipase D1. Sci. Signal. 3, ra1 (2010). (2015).

REVIEWS

13. Liu, S. T. et al. Synaptic vesicle-like lipidome of human 30. O’Reilly, M. C., Scott, S. A., Brown, H. A. & 54. Rudolph, A. E. et al. Expression, characterization, and cytomegalovirus virions reveals a role for SNARE Lindsley, C. W. Further evaluation of novel structural mutagenesis of the Yersinia pestis murine toxin, machinery in virion egress. Proc. Natl Acad. Sci. USA modifications to scaffolds that engender PLD isoform a phospholipase D. J. Biol. Chem. 274, 11824–11831 108, 12869–12874 (2011). selective inhibition. Bioorg. Med. Chem. Lett. 24, (1999). 14. Oguin, T. H. et al. Phospholipase D facilitates efficient 5553–5557 (2014). 55. Hinnebusch, B. J. et al. Role of Yersinia murine toxin in entry of influenza virus, allowing escape from innate 31. Lindsley, C. W. et al. Allosteric Akt (PKB) inhibitors: survival of Yersinia pestis in the midgut of the flea immune inhibition. J. Biol. Chem. 289, discovery and SAR of isozyme selective inhibitors. vector. Science 296, 733–735 (2002). 25405–25417 (2014). Bioorg. Med. Chem. Lett. 15, 761–764 (2005). 56. Edwards, J. L., Entz, D. D. & Apicella, M. A. This paper demonstrates that viruses can activate 32. DeFeo-Jones, D. et al. Tumor cell sensitization to Gonococcal phospholipase D modulates the PLD enzymes in the host cell as part of infection. apoptotic stimuli by selective inhibition of specific Akt/ expression and function of complement receptor 3 in 15. Taylor, H. E. et al. Phospholipase D1 couples CD4+ PKB family members. Mol. Cancer Ther. 4, 271–279 primary cervical epithelial cells. Infect. Immun. 71, T cell activation to c‑Myc-dependent (2005). 6381–6391 (2003). deoxyribonucleotide pool expansion and HIV‑1 33. Bornhop, D. J., Baksh, M., Finn, M. G. & Kussrow, A. 57. Edwards, J. L. & Apicella, M. A. Neisseria replication. PLoS Pathog. 11, e1004864 (2015). Backscattering interferometric analysis of membrane gonorrhoeae PLD directly interacts with Akt kinase 16. Scott, S. A. et al. Design of isoform-selective materials. US Patent 20130040306 (2013). upon infection of primary, human, cervical epithelial phospholipase D inhibitors that modulate cancer cell 34. Lavieri, R. R. Synthesis, Development and Biochemical cells. Cell Microbiol. 8, 1253–1257 (2006). invasiveness. Nat. Chem. Biol. 5, 108–117 (2009). Characterization of Small Molecule, Isoform-Selective 58. Taylor, L. D., Nelson, D. E., Dorward, D. W., This was the first paper to show that Phospholipase D Inhibitors and Photoactivatable Whitmire, W. M. & Caldwell, H. D. Biological isoform-selective inhibition of PLD enzymes with Probes. Ph.D. thesis, Vanderbilt Univ. (2014). characterization of Chlamydia trachomatis plasticity small-molecule ligands is possible. 35. Ganesan, R., Mahankali, M. Alter, G. & Gomez- zone MACPF domain family protein CT153. Infect. 17. Scott, S. A., Mathews, T. P., Ivanova, P. T., Cambronero, J. Two sites of action for PLD2 inhibitors: Immun. 78, 2691–2699 (2010). Lindsley, C. W. & Brown, H. A. Chemical modulation of the enzyme catalytic center and an allosteric, 59. Schroeder, G. N. et al. Legionella pneumophila glycerolipid signaling and metabolic pathways. phosphoinositide binding pocket. Biochim. Biophys. effector LpdA is a palmitoylated phospholipase D Biochim. Biophys. Acta. 1841, 1060–1084 (2014). Acta. 1851, 261–272 (2015). virulence factor. Infect. Immun. 83, 3898–4002 18. Scott, S. A. et al. Discovery of desketoraloxifene 36. Bruntz, R. C., Taylor, H. E., Lindsley, C. W. & Brown, H. A. (2015). analogs as inhibitors of mammalian, Pseudomonas Phospholipase D2 mediates survival signaling through 60. Jacobs, A. C. et al. Inactivation of phospholipase D aeruginosa, and NAPE phospholipase D enzymes. direct regulation of Akt in glioblastoma cells. J. Biol. diminishes Acinetobacter baumannii pathogenesis. ACS Chem. Biol. 10, 421–432 (2015). Chem. 289, 600–616 (2014). Infect. Immun. 78, 1952–1962 (2010). 19. Oh, M. K., Yang, H. & Roberts, M. F. Using O-(n‑alkyl)- This study establishes a link between PLD2 activity 61. Russell, A. B. et al. Diverse type VI secretion N-(N,N’-dimethylethyl)phosphoramidates to and AKT in survival pathways that are crucial in phospholipases are functionally plastic antibacterial investigate the role of Ca2+ and interfacial binding in cancer. effectors. Nature 496, 508–512 (2013). a bacterial phospholipase D. Biochim. Biophys. Acta. 37. Cummings, S. R. et al. The effect of raloxifene on risk of 62. Wilderman, P. J., Vasil, A. I., Martin, W. E., 1649, 146–153 (2003). breast cancer in postmenopausal women: results from Murphy, R. C. & Vasil, M. L. Pseudomonas aeruginosa 20. Egertová, M., Simon, G. M., Cravatt, B. F. & the MORE randomized trial. Multiple Outcomes of synthesizes phosphatidylcholine by use of the Elphick, M. R. Localization of N‑acyl Raloxifene Evaluation. JAMA 281, 2189–2197 (1999). phosphatidylcholine synthase pathway. J. Bacteriol. phosphatidylethanolamine phospholipase D (NAPE- 38. Plowman, P. N. Tamoxifen as adjuvant therapy in breast 184, 4792–4799 (2002). PLD) expression in mouse brain: a new perspective on cancer. Current status. Drugs 46, 819–833 (1993). This paper reports the identification of a novel N‑acylethanolamines as neural signaling molecules. 39. Eisen, S. F. & Brown, H. A. Selective estrogen receptor bacterial virulence factor. J. Comp. Neurol. 506, 604–615 (2008). modulators differentially regulate phospholipase D 63. Spencer, C. & Brown, H. A. Biochemical characterization 21. Monovich, L. et al. Optimization of halopemide for catalytic activity in ER‑negative breast cancer cells. of Pseudomonas aeruginosa phospholipase D. phospholipase D2 inhibition. Bioorg. Med. Chem. Lett. Mol. Pharmacol. 62, 911–920 (2002). Biochemistry 54, 1208–1218 (2015). 17, 2310–2311 (2007). This is the first study to report that drugs used in 64. Saïdani, N. et al. Discovery of compounds blocking the This report from Novartis catalysed current humans directly inhibit PLD activity. proliferation of Toxoplasma gondii and Plasmodium research into PLD inhibitors by showing that 40. Grese, T. A. et al. Molecular determinants of tissue falciparum in a chemical space based on piperidinyl- halopemide, an antipsychotic agent used in selectivity in estrogen receptor modulators. Proc. benzimidazolone analogs. Antimicrob. Agents humans, is also an inhibitor of PLD2. Natl Acad. Sci. USA 94, 14105–14110 (1997). Chemother. 58, 2586–2597 (2014). 22. Loonen, A. J., Soe-Agnie, C. J. & Soudijn, W. Effects of 41. Sun, X. & Whittaker, G. R. Entry of influenza virus. 65. Gomez-Cambronero, J. Phospholipase D in cell halopemide on GABA receptor binding, uptake and Adv. Exp. Med. Biol. 790, 72–82 (2013). signaling: from myriad of cell functions to cancer growth release. Brain Res. 210, 485–492 (1981). 42. Blaising, J. et al. Silibinin inhibits hepatitis C virus and metastasis. J. Biol. Chem. 289, 22557–22566 23. Loonen, A. J. & Soudijn, W. Halopemide, a new entry into hepatocytes by hindering clathrin- (2014). psychotropic agent. Cerebral distribution and dependent trafficking. Cell. Microbiol. 15, 66. Zhang, Y. & Frohman, M. A. Cellular and physiological receptor interactions. Pharm. Weekbl. Sci. 7, 1–9 1866–1882 (2013). roles of phospholipase D1 in cancer. J. Biol. Chem. (1985). 43. Kolokoltsov, A. A. et al. Small interfering RNA 289, 22567–22574 (2014). 24. Su, W. et al. 5‑Fluoro‑2‑indolyl des-chlorohalopemide profiling reveals key role of clathrin-mediated 67. Kang, D. W., Choi, K. Y. & Min do, S. Functional (FIPI), a phospholipase D pharmacological inhibitor endocytosis and early endosomes formation for regulation of phospholipase D expression in cancer that alters cell spreading and inhibits chemotaxis. infection by respiratory syncytial virus. J. Virol. 81, and inflammation. J. Biol. Chem. 289, 22575–22582 Mol. Pharmacol. 75, 437–446 (2009). 7786–7800 (2007). (2014). 25. Lewis, J. A. et al. Design and synthesis of isoform- 44. Burkard, C. et al. Coronavirus cell entry occurs through 68. Foster, D. A., Salloun, D., Mennon, D. & Frias, M. A. selective phospholipase D (PLD) inhibitors. Part I: the endo-/lysosomal pathway in a proteolysis-dependent Phospholipase D and the maintenance of phosphatidic impact of alternative halogenated privileged manner. PLoS Pathog. 10, e1004502 (2014). acid levels for regulation of mammalian target of structures for PLD1 specificity. Bioorg. Med. Chem. 45. Yang, L. et al. Role of phospholipase D2/phosphatidic rapamycin (mTOR). J. Biol. Chem. 289, 22583–22588 Lett. 19, 1916–1920 (2009). acid signal transduction in μ- and δ-opioid receptor (2014). This paper demonstrates that highly selective endocytosis. Mol. Pharmacol. 78, 105–112 (2010). 69. Uchida, N., Okamura, S., Nagamachi, Y. & PLD1 inhibitors can be developed. 46. Kang, D. W. et al. Autoregulation of phospholipase D Yamashita, S. Increased phospholipase D activity in 26. Lavieri, R. et al. Design and synthesis of isoform- activity is coupled to selective induction of phospholipase human breast cancer. J. Cancer Res. Clin. Oncol. 123, selective phospholipase D (PLD) inhibitors. Part II: D1 expression to promote invasion of breast cancer 280–285 (1997). identification of the 1,3,8‑triazaspiro[4,5]decan‑4‑one cells. Int. J. Cancer 128, 805–816 (2011). 70. Noh, D. Y. et al. Overexpression of phospholipase D1 privileged structure that engenders PLD2 selectivity. 47. Wang, R. et al. The transcription factor Myc controls in human breast cancer tissues. Cancer Lett. 161, Bioorg. Med. Chem. Lett. 19, 2240–2243 (2009). metabolic reprogramming upon T lymphocyte 207–214 (2000). 27. Lavieri, R. R. et al. Design, synthesis and biological activation. Immunity 35, 871–882 (2011). 71. Chen, Y., Zheng, Y. & Foster, D. A. Phospholipase D evaluation of halogenated N-(2-(4‑oxo‑1‑phenyl‑1,3,8‑tri 48. Korin, Y. D. & Zack, J. A. Nonproductive human confers rapamycin resistance in human breast cancer azasprio[4.5]decan‑8‑yl)ethyl)benzamides: discovery of immunodeficiency virus type 1 infection in nucleoside- cells. Oncogene 22, 3937–3942 (2003). an isoform-selective small molecule phospholipase D2 treated G0 lymphocytes. J. Virol. 73, 6526–6532 (1999). 72. Eder, A. M., Sasagawa, T., Mao, M., Aoki, J. & Mills, G. B. (PLD2) inhibitor. J. Med. Chem. 53, 6706–6719 (2010). 49. Plesa, G. et al. Addition of deoxynucleosides enhances Constitutive and lysophosphatidic acid (LPA)-induced LPA This paper demonstrates that highly selective human immunodeficiency virus type 1 integration and production role of phospholipase D and phospholipase PLD2 inhibitors can be developed. 2LTR formation in resting CD4+ T cells. J. Virol. 81, A2. Clin. Cancer Res.6, 2482–2491 (2000). 28. O’Reilly, M. C. et al. Development of dual PLD1/2 and 13938–13942 (2007). 73. Iorio, E. et al. Activation of phosphatidylcholine cycle PLD2 selective inhibitors from a common 50. Sturdevant, C. B. et al. Compartmentalized replication enzymes in human epithelial ovarian cancer cells. 1,3,8‑triazaspiro[4.5]decane core: discovery of of R5 T cell-tropic HIV‑1 in the central nervous system Cancer Res. 70, 2126–2135 (2010). ML298 and ML299 that decrease invasive migration early in the course of infection. PLoS Pathog. 11, 74. Hsu, Y. L. et al. Phospholipase D signaling pathway is in U87‑MG glioblastoma cells. J. Med. Chem. 56, e1004720 (2015). involved in lung cancer-derived IL‑8 increased 2695–2699 (2013). 51. Hyodo, K. et al. Phosphatidic acid produced by osteoclastogenesis. Carcinogenesis 31, 587–596 29. O’Reilly, M. C. et al. Discovery of a highly selective phospholipase D promotes RNA replication of a plant (2010). PLD2 inhibitor, VU0468809 (ML395): a new probe RNA virus. PLoS Pathog. 11, e1004909 (2015). 75. Yoshida, M., Okamura, S., Kodaki, T., Mori, M. & with improved physiochemical properties and broad 52. Mathews, T. P. et al. Human phospholipase D activity Yamashita, S. Enhanced levels of oleate- dependent spectrum antiviral activity against influenza strains. transiently regulates pyrimidine biosynthesis in malignant and Arf-dependent phospholipase D isoforms in ChemMedChem. 9, 2633–2637 (2014). gliomas. ACS Chem. Biol. 10, 1258–1268 (2015). experimental colon cancer. Oncol. Res. 10, 399–406 This study reports the seminal finding that 53. Ajl, S. J., Reedal, J. S., Durrum, E. L. & Warren, J. J. (1998). selective PLD2 inhibition is efficacious across Studies on plague. I. Purification and properties of the 76. Yamada, Y. et al. Association of a polymorphism of the a broad panel of influenza viruses, including toxin of Pasteurella pestis. J. Bacteriol. 70, 158–169 phospholipase D2 gene with the prevalence of pandemic strains. (1955). colorectal cancer. Mol. Med. 81, 126–131 (2003).

REVIEWS

77. Zhao, Y. et al. Increased activity and intranuclear 94. Cruchaga, C. et al. Rare coding variants in the 109. Lambeth, J. D. et al. ADP-ribosylation factor expression of phospholipase D2 in human renal cancer. phospholipase D3 gene confer risk for Alzheimer’s functions synergistically with a 50‑kDa cytosolic Biochem. Biophys. Res. Commun. 278, 140–143 disease. Nature 505, 550–554 (2014). factor in cell-free activation of human neutrophil (2000). 95. Ahn, M., Min, D. S., Kang, J., Jung, K. & Shin, T. phospholipase D. J. Biol. Chem. 270, 2431–2434 78. Douziech, N. et al. Growth effects of regulatory Increased expression of phospholipase D1 in the (1995). peptides and intracellular signaling routes in human spinal cords of rats with experimental autoimmune 110. Kwak, J. Y., Lopez, I., Uhlinger, D. J., Ryu, S. H. & pancreatic cancer cell lines. Endocrine. 9, 171–183 encephalomyelitis. Neurosci. Lett. 316, 95–98 Lambeth, J. D. RhoA and a cytosolic 50‑kDa factor (1998). (2001). reconstitute GTPγS‑dependent phospholipase D 79. Qi, C. et al. Lysophosphatidic acid stimulates 96. Lambert, J. C. et al. PLD3 and sporadic Alzheimer’s activity in human neutrophil subcellular fractions. phospholipase D activity and cell proliferation in PC‑3 disease risk. Nature 520, E1 (2015). J. Biol. Chem. 270, 27093–27098 (1995). human prostate cancer cells. J. Cell. Physiol. 174, 97. Van der Lee, S. J. et al. PLD3 variants in population 111. Park, S. K., Min, D. S. & Exton, J. H. Definition of the 261–272 (1998). studies. Nature 520, E2–E3 (2015). protein kinase C interaction site of phospholipase D. 80. Zhao, C., Du, G., Skowronek, K., Frohman, M. A. & 98. Hooli, B. V. et al. PLD3 gene variants and Alzheimer’s Biochim. Biophys. Res. Commun. 244, 364–367 Bar-Sagi, D. Phospholipase D2‑generated disease. Nature 520, E7–E8 (2015). (1998). phosphatidic acid couples EGFR stimulation to Ras 99. Tramutola, A. et al. Alteration of mTOR signaling 112. Exton, J. H. Regulation of phospholipase D. Biochim. activation by SOS. Nat. Cell Biol. 9, 706–712 occurs early in the progression of Alzheimer disease Biophys. Acta. 1439, 121–136 (1999). (2007). (AD): analysis of brain from subjects with pre-clinical 113. Brown, H. A., Henage, L. G., Preininger, A. M., 81. Buchanan, F. G. et al. Requirement of phospholipase AD, amnestic mild cognitive impairment and late- Xiang, Y. & Exton, J. H. Biochemical analysis of D1 activity in H-RasV12-induced transformation. Proc. stage AD. J. Neurochem. 133, 739–749 phospholipase D. Methods Enzymol. 434, 49–87 Natl Acad. Sci. USA 102, 1638–1642 (2005). (2015). (2007). 82. Cipriano, R. et al. FAM83B mediates EGFR- and RAS- 100. Jang, Y. H., Choi, K. Y. & Min, D. S. Phospholipase 114. Honda, A. et al. Phosphatidylinositol 4‑phosphate driven oncogenic transformations. J. Clin. Invest. 122, D‑mediated autophagic regulation is a potential target 5‑kinase α is a downstream effector of the small 3197–3210 (2012). for cancer therapy. Cell Death Differ. 21, 533–546 G protein ARF6 in membrane ruffle formation. Cell 83. Cipriano, R. et al. Hyperactivation of EGFR and (2014). 99, 521–532 (1999). downstream effector phospholipase D1 by oncogenic 101. Ivanova, P. T. et al. Lipid composition of viral envelope 115. Kooijman, E. E., Chupin, V., de Kruijff, B. & FAM83B. Oncogene 33, 3298–3306 (2014). of three strains of influenza virus — not all viruses are Burger, K. N. Modulation of membrane curvature by 84. Sun, Y. & Chen, J. mTOR signaling: PLD takes center created equal. ACS Infect. Dis. 1, 399–452 (2015). phosphatidic acid and lysophosphatidic acid. Traffic. stage. Cell Cycle 7, 3118–3123 (2008). 102. Powner, D. J., Pettitt, T. R., Anderson, R., Nash, G. B. 4, 162–174 (2002). 85. Saito, M. & Kanfer, J. Solubilization and properties of & Wakelam, M. J. Stable adhesion and migration of 116. Kooijman, E. E. & Burger, K. N. Biophysics and a membrane-bound enzyme from rat brain catalyzing human neutrophils requires phospholipase function of phosphatidic acid: a molecular a base-exchange reaction. Biochem. Biophys. Res. D‑mediated activation of the integrin CD11b/CD18. perspective. Biochim. Biophys. Acta. 1791, 881–888 Commun. 53, 391–398 (1973). Mol. Immunol. 44, 3211–3221 (2007). (2009). 86. Klein, J., Chalifa, V., Liscovitch, M. & Löffelholz, K. 103. Brown, H. A., Gutowski, S., Moomaw, C. R., Role of phospholipase D activation in nervous system Slaughter, C. & Sternweis, P. C. ADP-ribosylation Acknowledgements physiology and pathophysiology. J. Neurochem. 65, factor (ARF), a small GTP-dependent regulatory The authors are grateful for generous support from the US 1445–1455 (1995). protein, stimulates phospholipase D activity. Cell 75, National Institutes of Health (NIGMS 1R01GM121629; HAB 87. Oliveira, T. G. & Di Paolo, G. Phospholipase D in brain 1137–1144 (1993). and HHSN272200800058C; H.A.B. and P.G.T.), the Centers function and Alzheimer’s disease. Biochim. Biophys. This paper identifies the first protein and lipid of Excellence for Influenza Research and Surveillance (CEIRS; Acta. 1801, 799–805 (2010). modulators of PLD using a novel in vitro HHSN272201400006C; P.G.T.) and the Molecular Libraries 88. Lindsley, C. W. & Brown, H. A. Phospholipase D as biochemical assay. Probe Production Centers Network (MLPCN; C.W.L.), in addi‑ a therapeutic target in brain disorders. 104. Brown, H. A., Gutowski, S., Kahn, R. A. & tion to Voices Against Brain Cancer (H.A.B.), the Vanderbilt Neuropsychopharmacology 37, 301–302 (2012). Sternweis, P. C. Partial purification and Institute of Chemical Biology (H.A.B.), the Vanderbilt 89. Jung, K. et al. Upregulation of phospholipase D1 in characterization of Arf-sensitive phospholipase D from University Bixler-Johnson-Mayes Endowed Chair (H.A.B.), the the spinal cords of rats with clip compression injury. porcine brain. J. Biol. Chem. 270, 14935–14943 William K. Warren Jr. Endowment (C.W.L.), and The Hartwell Neurosci. Lett. 336, 126–130 (2003). (1995). Foundation (P.G.T.). The authors thank T. H. Oguin 3rd for 90. Stegner, D. et al. Pharmacological inhibition of 105. Jiang, X., Gutowski, S., Singer, W. D. & helpful discussions, as well as M. D. Armstrong and P. T. phospholipase D protects mice from occlusive thrombus Sternweis, P. C. Assays and characterization of Ivanova for invaluable discussions and assistance with edito‑ formation and ischemic stroke — brief report. mammalian phosphatidylinositol‑4,5‑bisphosphate- rial and graphical content. Arterioscler. Thromb. Vasc. Biol. 33, 2212–2217 (2013). sensitive phospholipase D. Methods Enzymol. 345, 91. Jin, J.‑K. et al. Phospholipase D1 is up‑regulated in 328–334 (2002). Competing interests statement the mitochondrial fraction from the brains of 106. Morris, A. J., Engelbrecht, J. & Frohman, M. A. The authors declare competing interests: see Web version Alzheimer’s disease patients. Neurosci. Lett. 407, Structure and regulation of phospholipase D. Trends for details. 263–267 (2006). Pharmacol. Sci. 17, 182–185 (1996). 92. Jin, J.‑K. et al. Phospholipase D1 is associated with 107. Cockcroft, S. et al. Phospholipase D: a downstream amyloid precursor protein in Alzheimer’s disease. effector of ARF in granulocytes. Science 263, FURTHER INFORMATION Neurobiol. Aging 28, 1015–1027 (2007). 523–526 (1994). KEGG enzyme commission number 3.1.4.4: 93. Cai, D. et al. Presenilin‑1 uses phospholipase D as 108. Brown, H. A. & Sternweis, P. C. Stimulation of http://www.genome.jp/dbget-bin/www_bget?ec:3.1.4.4 a negative regulator of β-amyloid formation. Proc. phospholipase D by ADP-ribosylation factor. ALL LINKS ARE ACTIVE IN THE ONLINE PDF Natl Acad. Sci. USA 103, 1941–1946 (2006). Methods Enzymol. 257, 313–324 (1995).